DecisionBox Enterprise bets air-gapped is the only serious enterprise agent pitch
DecisionBox just shipped an Enterprise SKU of their SQL agent and the positioning is the news. Air-gapped deployment, self-hosted models via Ollama, zero bytes leave your network, AGPL v3 open-source core with a plugin architecture on top. This is the first agent product of the week that openly says the cloud version is not the serious offering.
The core product is what it has always been, an autonomous agent that writes SQL against your warehouse and ships validated findings. You do not ask it specific questions. It figures out what is worth surfacing, runs the query, and sends a report. What changed is the deployment story, three-layer data governance with schema filtering, column redaction, and query rewriting, plus SSO, RBAC, complete audit logs, Slack integration. The kind of feature checklist an enterprise procurement team wrote, not a product manager.
The real editorial point is that the air-gapped-agent category just became a viable GTM wedge. For eighteen months the default assumption was that enterprise AI meant sending your warehouse schema to OpenAI or Anthropic. Banks and hospitals rejected that out of hand. The plays that could have won, local LLMs fine-tuned on your schema, plugin architectures that let you swap the model, no data egress at all, were deemed commercially unsexy. DecisionBox is making the opposite bet, the companies actually signing seven-figure contracts are the ones who cannot send data out, and the agent that fits that shape wins them all.
This is also the third PH launch from DecisionBox, April 7 and April 14 were earlier chapters, April 22 is the Enterprise wedge. The cadence is instructive, ship the cloud version first to prove the agent, then package the air-gapped story for people who would never have signed up for cloud. If the model holds, expect more open-core agent companies to follow the same three-beat launch in the next six months.
Free two-week POC at https://www.producthunt.com/products/decisionbox.
← Back to all articles
The core product is what it has always been, an autonomous agent that writes SQL against your warehouse and ships validated findings. You do not ask it specific questions. It figures out what is worth surfacing, runs the query, and sends a report. What changed is the deployment story, three-layer data governance with schema filtering, column redaction, and query rewriting, plus SSO, RBAC, complete audit logs, Slack integration. The kind of feature checklist an enterprise procurement team wrote, not a product manager.
The real editorial point is that the air-gapped-agent category just became a viable GTM wedge. For eighteen months the default assumption was that enterprise AI meant sending your warehouse schema to OpenAI or Anthropic. Banks and hospitals rejected that out of hand. The plays that could have won, local LLMs fine-tuned on your schema, plugin architectures that let you swap the model, no data egress at all, were deemed commercially unsexy. DecisionBox is making the opposite bet, the companies actually signing seven-figure contracts are the ones who cannot send data out, and the agent that fits that shape wins them all.
This is also the third PH launch from DecisionBox, April 7 and April 14 were earlier chapters, April 22 is the Enterprise wedge. The cadence is instructive, ship the cloud version first to prove the agent, then package the air-gapped story for people who would never have signed up for cloud. If the model holds, expect more open-core agent companies to follow the same three-beat launch in the next six months.
Free two-week POC at https://www.producthunt.com/products/decisionbox.
Comments